1 On the Seasonality of Arctic Haze
∗ 2 Zhaoyi Shen
3 Program in Atmospheric and Oceanic Sciences, Princeton University, Princeton, New Jersey.
4 Yi Ming, Larry W. Horowitz, V. Ramaswamy and Meiyun Lin
5 NOAA/Geophysical Fluid Dynamics Laboratory, Princeton, New Jersey.
∗ 6 Corresponding author address: Zhaoyi Shen, NOAA Geophysical Fluid Dynamics Laboratory,
7 201 Forrestal Rd., Princeton, NJ 08540.
8 E-mail: [email protected]
Generated using v4.3.2 of the AMS LATEX template 1 ABSTRACT
9 Arctic haze has a distinct seasonality with peak concentrations in winter but
10 pristine conditions in summer. It is demonstrated that the Geophysical Fluid
11 Dynamics Laboratory (GFDL) atmospheric general circulation model AM3
12 can reproduce the observed seasonality of Arctic black carbon (BC), an im-
13 portant component of Arctic haze. We use the model to study how large-scale
14 circulation and removal drive the seasonal cycle of Arctic BC. It is found that
15 despite large seasonal shifts in the general circulation pattern, the transport
16 of BC into the Arctic varies little throughout the year. The seasonal cycle of
17 Arctic BC is attributed mostly to variations in the controlling factors of wet
18 removal, namely the hydrophilic fraction of BC and wet deposition efficiency
19 of hydrophilic BC. Specifically, a confluence of low hydrophilic fraction and
20 weak wet deposition, owing to slower aging process and less efficient mixed-
21 phase cloud scavenging, respectively, is responsible for the wintertime peak
22 of BC. The transition to low BC in summer is the consequence of a gradual
23 increase in the wet deposition efficiency, while the increase of BC in late fall
24 can be explained by a sharp decrease in the hydrophilic fraction. The results
25 presented here suggest that future changes in the aging and wet deposition
26 processes can potentially alter the concentrations of Arctic aerosols and their
27 climate effects.
2 28 1. Introduction
29 The accumulation of visibility-reducing aerosols in the Arctic during late winter and early spring
30 (known as Arctic haze) was first discovered in the 1950s (Mitchell 1957). The haze has its root
31 cause in the long-range transport of air pollution originating from the mid-latitude industrial re-
32 gions (Barrie 1986), and can have an influence on Arctic climate (Law and Stohl 2007). The haze
33 is a mixture of both light-scattering and light-absorbing aerosols. The aerosols pose strong radia-
34 tive perturbations by scattering and/or absorbing solar radiation, by interacting with clouds, and
35 by reducing the surface albedo when deposited onto snow and ice (Quinn et al. 2007). The surface
36 temperature in the Arctic increased more than the global average since the late 20th century, coin-
37 ciding with a rapid decline of sea ice (Bindoff et al. 2013). Besides greenhouse gases, decreased
38 (increased) scattering (absorbing) aerosols were postulated to have contributed to this amplified
39 Arctic climate change (Shindell and Faluvegi 2009).
40 Sulfate and black carbon (BC) are the dominant scattering and absorbing aerosols in the Arctic,
41 respectively (Law and Stohl 2007). Distinct seasonal cycles of Arctic sulfate and BC concentra-
42 tions are present in measurements. Surface observations at Alert and Barrow show that BC con-
43 centrations tend to peak during late winter and early spring before starting to decline in April, and
44 reach a minimum during summer (Sharma et al. 2006). Aircraft measurements of sulfate reveal
45 similar seasonal variations, with aerosol layers at higher altitudes persisting into May (Scheuer
46 2003).
47 Different hypotheses have been posited to explain the seasonal cycle of Arctic haze. Previous
48 studies have shown that European emissions dominate Arctic aerosols, with smaller contributions
49 from East Asia and North America (e.g., Stohl 2006; Shindell et al. 2008). A dynamically oriented
50 view holds that during the haze season (winter and spring), meridional transport from mid-latitude
3 51 source regions to the Arctic is stronger due to vigorous large-scale circulation. The presence of
52 Siberian high pressure helps steer polluted European air into the Arctic by transient and stationary
53 eddies (Barrie 1986; Iversen and Joranger 1985). The diabatic cooling of air traveling over ice and
54 snow also facilitates transport to the Arctic lower troposphere. In contrast, pollution is diabatically
55 transported to higher altitudes and diluted in summer (Klonecki et al. 2003).
56 A competing theory is that Arctic haze occurs in winter owing to slower removal. As a major
57 sink term for aerosols, wet scavenging by precipitation (rain and snow) is modulated heavily by
58 cloud microphysics. Aerosols are not effectively removed by ice clouds since soluble aerosols
59 such as sulfate are generally poor ice nuclei (IN). BC particles become effective IN only when
60 the temperature is below ∼240 K (Friedman et al. 2011). In mixed-phase clouds, the Bergeron
61 process (i.e. evaporation of liquid droplets in the presence of ice crystals) releases aerosols con-
62 tained in cloud droplets back into air (Cozic et al. 2008), so the wet scavenging in mixed-phase
63 clouds is much less efficient than in liquid clouds (Liu et al. 2011; Browse et al. 2012). The low
64 efficiency of ice cloud and mixed-phase cloud scavenging favors accumulation of aerosols at cold
65 temperatures. Dry deposition also has seasonal variations and is weaker in winter when the stable
66 boundary layer inhibits turbulent mixing (Quinn et al. 2007). It, however, accounts for only a small
67 portion of the total removal and affects mainly the surface concentrations of Arctic haze (Liu et al.
68 2011). Another key factor unique to BC is the aging process, which refers to the transformation
69 from hydrophobic to hydrophilic aerosols due to coating by soluble species (Petters et al. 2006).
70 Only aged BC particles can act as cloud condensation nuclei (CCN) and be removed by in-cloud
71 scavenging. As a result, the aging rate has a large effect on global BC concentrations and distribu-
72 tions. Experimental studies have found that condensation of gaseous sulfuric acid (H2SO4) is an
73 effective aging mechanism for BC particles (Zhang et al. 2008); reduced concentrations of H2SO4
4 74 due to slower oxidation of SO2 by the hydroxyl radical (OH) in winter results in a lower aging rate
75 and thus weaker wet deposition (Liu et al. 2011).
76 It is important to note that the aforementioned mechanisms (namely large-scale circulation,
77 cloud microphysics and aging) are not mutually exclusive; they could all act to induce season-
78 ality. Yet, the relative importance of these mechanisms in shaping the pronounced seasonal cycle
79 of Arctic haze remains unclear. While it is hard to separate these factors cleanly using observations
80 alone, we approach the issue here using a comprehensive global model. In this study we apply
81 an atmospheric general circulation model (AGCM) to analyze the factors controlling the seasonal
82 cycle of Arctic haze with a focus on BC, an important constituent of Arctic haze with a potentially
83 important influence on Arctic climate.
84 2. Model Description
85 This study uses the Geophysical Fluid Dynamics Laboratory (GFDL) AM3 AGCM (Donner
86 et al. 2011) with a cubed-sphere grid resolution of ∼100 km and 48 hybrid vertical levels from
87 the surface to ∼1 Pa. We conduct a six-year hindcast simulation (2008-2013), following one year
88 of spin-up. The emission inventories reflect 2008-2013 conditions. Anthropogenic emissions of
89 aerosol and ozone precursors with seasonal variations are based on Hemispheric Transport of Air
90 Pollution (HTAP) v2 - a mosaic of regional and global emission inventories for the years 2008 and
91 2010 (Janssens-Maenhout et al. 2015), and are held constant after 2010. Daily-resolving biomass
92 burning emissions for 2008-2013 are adopted from the Fire Inventory from National Center for
93 Atmospheric Research (NCAR) (Wiedinmyer et al. 2011) and emitted in the model surface layer.
94 The model is forced with observed sea surface temperatures and sea ice, and horizontal winds are
95 nudged to the reanalyses from the National Centers for Environmental Prediction (NCEP) Global
◦ ◦ 96 Forecasting System at approximately 1.4 ×1.4 horizontal resolution using a pressure-dependent
5 97 nudging technique (Lin et al. 2012). The latter makes it possible to compare model simulations
98 with observations for specific flight campaigns.
99 The treatment of BC in AM3 used in this study differs from that described by Donner et al.
100 (2011), and has been discussed extensively by Liu et al. (2011) and Fan et al. (2012); here we
101 summarize briefly the key features. AM3 includes two types of BC: hydrophobic (BCpo) and
102 hydrophilic (BCpi). 80% (40%) of BC emitted from anthropogenic (biomass burning) sources
103 is assumed to be hydrophobic. The hydrophobic BC is then converted to the hydrophilic form
104 at a variable aging rate (Liu et al. 2011), which depends on condensation of sulfuric acid (the
105 rate of which is assumed to be proportional to OH concentrations) and other processes such as
106 coagulation (represented by adding a small constant term to the aging rate coefficient, see Section
107 6 for more discussion). Only hydrophilic BC can be removed by in-cloud scavenging, which is
−1 108 parameterized using a first-order rate coefficient (kscav, s ) (Fan et al. 2012):
Fscav,1Prain + Fscav,2(1 − fberg)Psnow + Fscav,3 fbergPsnow kscav = , (1) Qliq + Qice
−3 −1 109 where Prain and Psnow are the 3-dimensional rain and snow rates (kg m s ), respectively, and
−3 110 Qliq and Qice the liquid and ice cloud water contents (kg m ), respectively. Fscav,i is the scaveng-
111 ing efficiency (i.e. the fraction of BC that is incorporated into cloud droplets or ice crystals and
112 removed by precipitation) for precipitation type i (i =1, 2, 3). fberg is the fraction of snow produced
113 by the Bergeron process. In this study, Fscav,1 and Fscav,2 are set to 0.2, and Fscav,3 is set to 0.01
114 to account for the less efficient removal of aerosols by snow produced by the Bergeron process
115 than by rain and snow produced by riming and homogeneous freezing. Both hydrophobic and hy-
116 drophilic BC can be removed by below-cloud scavenging and dry deposition. The dry deposition
117 velocity is calculated using the empirical resistance-in-series method with a surface-dependent
6 118 collection efficiency (Gallagher 2002), resulting in a much smaller dry deposition velocity over
119 snow and ice than over land surfaces (soils and canopy).
120 3. Simulated and observed seasonal cycle of Arctic BC
121 Simulating Arctic BC, especially its seasonal cycle, remains a challenge for the current gener-
122 ation of models. Many models underestimate Arctic BC concentrations in winter by more than
123 an order of magnitude and thus fail to reproduce the observed seasonal variations (Shindell et al.
124 2008). Figure 1 shows the AM3 simulated and observed monthly mean BC surface concentrations
125 at Alert, Barrow, and Zeppelin (Eleftheriadis et al. 2009; Sharma 2004; Sharma et al. 2006). Mea-
126 sured BC exhibits similar seasonal variations at all three stations. Its concentrations peak in winter
127 or early spring, followed by a rather precipitous decrease in April and May. The lowest concentra-
128 tions occur from June to October. On average, BC concentrations in winter (DJF) are higher than
129 in summer (JJA) by a factor of 3-4. The model is able to capture the seasonal cycle, but appears
130 to underestimate BC at Alert and Barrow by a factor of 2-3 throughout the year. This discrepancy
131 may result from model deficiencies, including uncertainties in BC emissions and deposition, or
132 from observational errors, as some of the measurements are indirect and may be subject to rather
133 large biases (Bond et al. 2013).
◦ ◦ 134 Figure 1 also compares the simulated BC vertical profiles at high latitudes (66 -85 N) with the
135 aircraft measurements made during the HIAPER Pole-to-Pole Observations (HIPPO) campaigns
136 (Schwarz et al. 2013; Wofsy 2011). The seasonal variations are evident in the data. In January,
137 high concentrations of BC are confined within the boundary layer (HIPPO1). During early spring,
138 there are enhancements of BC at higher altitudes (HIPPO3). BC concentrations at all altitudes start
139 to decrease in June, and remain low throughout summer (HIPPO4 and HIPPO5). Despite overes-
7 140 timates in the free troposphere for some flights, the model generally performs well in simulating
141 BC vertical profiles during all seasons.
142 Based on the above comparisons, we conclude that the model is capable of simulating the sea-
143 sonal cycle of Arctic BC, and thus can be used to study its underlying mechanisms. This improve-
144 ment in AM3 simulated Arctic BC is attributed to the modified BC-related processes (aging, wet
145 removal and dry deposition) in the model, and Liu et al. (2011) discussed the sensitivity of BC
146 simulation to each process in detail. Since observations show similar seasonality of Arctic BC at
147 the surface and in the free troposphere, we will focus on the BC column burden averaged over the
148 Arctic for the rest of the paper. This allows us to take full advantage of the model and generalize
149 the results to the entire Arctic.
150 4. Controlling factors of Arctic BC
151 We employ a box model of the Arctic region to quantify the key factors controlling Arctic BC
152 concentrations. Given the relatively small emissions from local sources, the prevailing balance is
153 between the meridional BC transport into the Arctic and local deposition. One can write the rate
−2 154 of change in the average BC column burden (C, kg m ) as:
dC F = −W − D, (2) dt S
−1 2 155 where F is the total meridional flux into the Arctic (kg s ), S the Arctic surface area (m ), and
−2 −1 156 W and D the average wet and dry deposition rates (kg m s ), respectively. Since the dry
157 deposition of BC and the below-cloud scavenging of BCpo are small (less than 10% of the total
158 BC deposition in the model simulation) and can be neglected, Eq. (2) can be simplified as:
dC F = −W , (3) dt S pi
8 −2 −1 159 where Wpi is the average BCpi wet deposition rate (kg m s ). Wpi can be written as r · w ·C,
160 where r represents the dimensionless hydrophilic fraction of BC (Cpi/C, Cpi being the average
−1 161 BCpi column burden) and w represents the wet deposition efficiency of BCpi (Wpi/Cpi, s ),
162 which can be thought of as the BC concentration-weighted in-cloud scavenging rate coefficient
163 (kscav) defined in Eq. (1), and is different from the wet scavenging efficiency (Fscav,i). As the
164 AM3 simulated residence time of Arctic BC ranges from 6-20 days depending on the season (not
165 shown), we assume steady state on the monthly time scale and arrive at an expression for C:
F C = . (4) S · r · w
166 An inspection of Eq. (4) suggests that elevated BC concentrations could result from stronger
167 transport from mid-latitude source regions and/or weaker wet removal. General circulation pat-
168 terns are important for determining long-rage transport fluxes. Wet removal is reduced when a
169 smaller fraction of BC is hydrophilic or the wet deposition efficiency of BCpi is lower. The aging
170 process exerts a strong control over the hydrophilic fraction. The wet deposition efficiency is af-
171 fected mainly by the phase of precipitation because in-cloud scavenging efficiency differs among
172 liquid, ice and mixed-phase clouds.
173 We apply the box model to quantify the roles of the three main variables, namely the meridional
174 BC flux (F), BC hydrophilic fraction (r) and BCpi wet deposition efficiency (w), in controlling
175 the seasonality of Arctic BC. The monthly mean values of F, r and w averaged over the Arctic
◦ 176 (defined as poleward of 66 N) are computed from our AM3 model simulation. Figure 2a compares
177 the monthly mean BC column burdens calculated using the box model with AM3 simulations. The
178 good agreement validates the assumptions made in deriving Eq. (4). The box model captures the
179 seasonal cycle, suggesting that one can rationalize the seasonality of Arctic BC by examining the
180 three variables defined above.
9 181 5. Meridional transport
182 Like other anthropogenic aerosol species, BC is emitted mainly at the mid-latitude industrial
183 regions and carried into the Arctic by atmospheric transport. Isentropic airflow facilitates high-
184 level transport from warm and humid (high equivalent potential temperature, θe) areas such as
185 North America and East Asia, and low-level transport from comparatively low θe areas such as
186 Europe (Stohl 2006). Cross-isentropic transport due to diabatic heating or cooling also plays
187 an important role (Klonecki et al. 2003). The total meridional BC flux (F) can be decomposed
188 into contributions from the mean meridional circulation (MMC), stationary eddies, and transient
189 eddies:
{vc} = {[v][c]} + {v∗c∗} + {v0c0}, (5) | {z } | {z } | {z } MMC stationary eddies transient eddies
−1 −1 190 where v is the meridional wind velocity (m s ), and c the BC mass mixing ratio (kg kg ).
191 Overbars denote monthly means, square brackets zonal means, asterisks deviations from zonal
192 means, primes deviations from monthly means, and curly brackets zonal and vertical integrals
193 (from the surface to ∼100 hPa). Figure 2b shows the annual cycle of the total monthly mean
◦ 194 meridional BC flux into the Arctic (at 66 N) and its three components. Generally, the total BC
195 flux does not show much variation with time. The average flux is only ∼20% higher in DJF than
196 in JJA, far from sufficient to account for the column burden difference between the two seasons
197 (Figure 2a). The contribution of the mean meridional circulation is very small. The transport is
198 realized almost entirely by the eddy components. The flux by stationary eddies is comparable to
199 that by transient eddies in DJF, while the latter dominates in JJA.
200 Although the vertically integrated BC flux changes little throughout the year, its vertical structure
201 does vary with the season (Figure 3). The flux in DJF is characterized by two peaks (one in the
202 boundary layer, and the other at about 500 hPa), which are of different origins. In winter the polar
10 ◦ 203 dome (surface of constant potential temperatures) extends to about 40 N, allowing the low-level
204 transport of BC from Europe (Stohl 2006). The diabatic cooling occurring when relatively warm
205 air travels over a cold surface (i.e. strong inversion) keeps European BC in the boundary layer.
206 The flux at about 500 hPa is more likely to be a result of transport from lower-latitude regions such
207 as East Asia and North America. In JJA when BC from all regions experiences diabatic heating
208 and wet removal caused by precipitation, the flux has only one notable peak at about 800 hPa
209 originating from anthropogenic emissions in Europe and boreal forest fires over Eurasia. BC from
210 North America and East Asia is more likely to be transported diabatically to higher altitudes, and
211 diluted and rained out along the path to the Arctic (Klonecki et al. 2003).
212 We further explore the meridional BC flux in frequency space by applying a time filter. Since
213 the mean meridional flux is negligible, {vncn} approximates BC transport carried out by eddies
214 with time scales greater than 2n days (Hall et al. 1994). Note that the subscript n denotes means
215 of consecutive non-overlapping n-day periods. Figure 4 shows the time-filtered BC flux into the
216 Arctic as a fraction of the total flux. The relative contributions from eddies of different frequencies
217 to the total BC transport are different in the two seasons. In DJF, about 90% of the BC flux is
218 realized by eddies with time scales longer than 10 days. Slightly more than 40% of the BC flux
219 arises from eddies that persist longer than 60 days. In contrast, eddies with time scales longer
220 than 10 days account for less than 50% of the JJA flux, and eddies that persist longer than 60 days
221 have very little contribution to the total flux. Thus, while synoptic eddies dominate in JJA, low
222 frequency eddies contribute substantially to the total transport in DJF.
223 The prominence of transient eddies in all seasons suggests that the long-range transport of Arctic
224 haze can be represented, to first order, as turbulent diffusion of mid-latitude sources, despite the
225 complexities at the process level (Shaw 1981). The idea of simplifying eddy transport as turbulent
226 diffusion is widely used in understanding the atmospheric transport of heat and potential vorticity
11 227 (Held 1999). Here we apply it to study tracer transport. For a local down-gradient diffusion pro-
228 cess, the vertically integrated meridional transient eddy BC flux can be assumed to be proportional
229 to the meridional gradient of the BC column burden:
∂ {v0c0} = −D {c}, (6) ∂y
230 where D is the turbulent diffusivity. It is clear from Figure 5 that there is a strong negative correla-
231 tion (r = −0.75) between the transient eddy flux and meridional gradient averaged at a number of
◦ ◦ 232 latitudes between 40 -66 N. The magnitude of the slope of the best linear fit with zero intercept
6 2 −1 233 (2.24 × 10 m s ) represents the eddy diffusivity, which is within the range of the estimated
234 values in previous studies (e.g., Bolin and Keeling 1963; Newell et al. 1969; Held 1999). Note
235 that this diffusivity does not change substantially with the season.
236 6. Hydrophilic fraction
237 Figure 6a shows the monthly mean hydrophilic fraction (r) of Arctic BC, which has a pro-
238 nounced seasonal cycle that can be attributed to the parameterized aging mechanisms. The hy-
239 drophilic fraction in DJF is only ∼40%, while almost all BC is hydrophilic in JJA. The hydrophilic
240 fraction of Arctic BC is very close to that of the meridional BC flux into the Arctic (Figure 6a), in-
241 dicating that the annual cycle is shaped mainly by the aging process along the long-range transport
242 path rather than locally in the Arctic. Figure 6b shows the monthly mean e-folding aging time of
243 BC (the inverse of the aging rate coefficient, computed as the average BCpo concentration divided
◦ ◦ 244 by the average conversion rate from BCpo to BCpi) at 40 -66 N and over the Arctic. During the
245 transport from mid-latitudes to the Arctic, the average aging time is much longer in DJF (∼10
246 days) than in JJA (∼1 day), resulting in a substantially lower hydrophilic fraction in winter. In
12 247 the Arctic, the seasonal variations of BC aging time are even greater, which further amplifies the
248 seasonal cycle of the hydrophilic fraction.
249 In the model the BC aging is assumed to result primarily from the condensation of H2SO4 onto
250 BC aerosol surface, a common process that has been examined extensively in observational and
251 experimental studies. H2SO4 is produced from the gas-phase oxidation of SO2 by the OH radical
252 and is rapidly converted to aerosol phase via nucleation or condensation onto existing particles.
253 The aging rate is thus assumed to be proportional to OH concentrations. As a result of enhanced
254 solar radiation and specific humidity, OH concentrations are much higher in JJA than in DJF,
255 resulting in more rapid aging by condensation in summer. The aging can also occur through other
256 processes (e.g. coagulation), which are believed to be slower and less important than condensation
257 during long-range transport (Oshima et al. 2009). Therefore they are assumed to have a fixed e-
258 folding time of 20 days, which is longer than that for the aging via condensation, and thus do not
259 contribute to the seasonal cycle of the hydrophilic fraction.
260 The change in the hydrophilic fraction also helps explain the change in Arctic BC during spring
261 and fall. From October to November, there is a sharp decrease in the hydrophilic fraction, re-
262 sulting in a rapid buildup of BC. Similarly, from March to April, the increase in the hydrophilic
263 fraction contributes to the decline in BC concentrations. The hydrophilic fraction, however, is
264 fairly constant from April to September, in contrast with the continuous decrease in BC starting
265 from April. The wet deposition efficiency plays an important role in driving BC changes during
266 this time period, as discussed in the next section.
267 7. BCpi wet deposition efficiency
268 Figure 6c shows the monthly mean wet deposition efficiency of BCpi (w) in the Arctic. The
269 wet deposition efficiency in JJA is ∼20% higher than in DJF, contributing to the lower BC in
13 270 JJA than in DJF. The magnitude of its seasonal cycle, however, is much weaker than that of the
271 hydrophilic fraction (r) (Figure 6a). Yet, the wet deposition efficiency increases continuously by a
272 factor of 2 from May to August, driving the transition from moderate BC burdens in late spring to
273 exceedingly low burdens in summer.
274 The wet deposition efficiency is largely controlled by the in-cloud scavenging rate coefficient
275 [kscav in Eq. (1)]. To better understand the factors determining kscav, we analyze the conversion rate
276 coefficients of cloud condensate to rain (krain) and snow (ksnow) through which in-cloud scavenging
277 occurs, which are defined as:
Prain Psnow krain = , ksnow = . (7) Qliq + Qice Qliq + Qice
278 According to Eq. (1), kscav and thus the wet deposition efficiency is determined by krain rather than
279 ksnow in areas such as the Arctic where most snow is produced by the Bergeron process (Fan et al.
280 2012), since the scavenging efficiency for rain is much larger than for snow produced by Bergeron
281 process. Figure 6d shows the annual cycle of AM3 simulated krain and ksnow in the Arctic. [Prain,
282 Psnow, Qliq and Qice are first averaged over the Arctic and then krain and ksnow are calculated using
283 Eq. (7)]. As the atmosphere becomes warmer and holds more water vapor from DJF to JJA, both
284 rainfall (Prain) and cloud water content (Qliq + Qice) increase, and krain is higher in JJA than in
285 DJF since rainfall increases more rapidly, consistent with the seasonal cycle of the wet deposition
286 efficiency. From May to August, krain increases by a factor of 2, which helps explain the two-fold
287 increase in the wet deposition efficiency. On the other hand, ksnow is much larger in DJF than in
288 JJA because cold temperature favors snow formation, which is opposite to the seasonal variations
289 in the wet deposition efficiency. Therefore one may expect that models without different removal
290 efficiencies for liquid and mixed-phase clouds cannot reproduce the seasonal cycle of the wet
291 deposition efficiency, and thus the seasonal cycle of Arctic BC concentrations.
14 292 8. Concluding remarks
293 It has long been recognized that aerosols from mid-latitude source regions undergo long-range
294 transport and accumulate in the Arctic during late winter and early spring. Here we apply the
295 GFDL AM3 model to analyze the key factors affecting the seasonal variations in Arctic BC. The
296 model is able to reproduce the observed Arctic BC concentrations and seasonality, with 3-4 times
297 higher values in DJF than in JJA. We find that the seasonal cycle of Arctic BC is caused mainly
298 by the seasonality of wet deposition, with a secondary contribution from the long-range transport
299 flux.
300 The transport of BC at mid- to high latitudes occurs mainly through stationary and transient
301 eddies, rather than through the mean meridional circulation. Stationary eddies account for ∼40%
302 of the total flux in DJF, while virtually all the transport is realized through transient eddies in JJA.
303 The vertical distribution of meridional BC transport also varies seasonally. The vertical profile of
304 the BC flux into the Arctic in DJF has two peaks (one in the boundary layer and the other in the
305 mid-troposphere), while the BC flux in JJA is concentrated in the lower troposphere. The total
306 meridional BC flux into the Arctic, however, changes little throughout the year despite the shift in
307 large-scale circulation.
308 The wet removal depends on both BC hydrophilic fraction and BCpi wet deposition efficiency.
309 The hydrophilic fraction is smaller in DJF than in JJA due to the slower BC aging along the
310 long-range transport path to the Arctic during winter. This difference in the hydrophilic fraction
311 plays a dominant role in the large difference in BC concentrations between DJF and JJA. The wet
312 deposition efficiency is lower in DJF mainly because snow produced in mixed-phase clouds is less
313 efficient in removing BC than rain. The decrease in BC concentrations from late spring to summer
15 314 is due to a gradual but steady increase in the wet deposition efficiency, while the return of BC in
315 late autumn is mainly caused by a sharp decrease in the hydrophilic fraction.
316 Our results are consistent with the observational analysis of Garrett et al. (2011), which argued
317 that some combination of dry deposition and wet scavenging drives the seasonal cycle of aerosols
318 at low altitudes in the North American Arctic. Here we show that the dominance of wet deposition
319 in determining BC seasonal cycle applies to the entire Arctic. We further explain the seasonality
320 of wet deposition in terms of aging and cloud microphysical processes. While being influenced
321 by complicated physical and chemical factors, these processes are parameterized in the model in
322 a relatively simple way. Measurements of Arctic BC mixing state at different times are required
323 for verifying the assumed OH dependence of the aging rate. Measurements of BC wet deposi-
324 tion and concentrations in rain and snow will be particularly useful for constraining the rain and
325 snow scavenging efficiencies. It should also be noted that besides transport and wet deposition,
326 large uncertainties remain in BC emission inventories, which have not been discussed in this paper.
327 High-latitude BC emissions from gas flaring and residential combustion, which are underestimated
328 or completely missing in most current inventories, have a large contribution to Arctic BC concen-
329 trations and are potentially crucial for affecting its seasonality (Stohl et al. 2013). However, as
330 most of these BC emissions remain close to the surface, their contributions to BC in the mid- and
331 upper troposphere in the Arctic are small (Stohl et al. 2013) and will not affect our conclusions,
332 which are based on the BC column burden. Although the analysis in this paper focuses on BC, the
333 findings should be generally applicable to other components of Arctic haze. For example, sulfate
334 has a similar seasonal cycle, but with peak concentrations in spring as opposed to winter (as is the
335 case for BC) (Shindell et al. 2008). The difference in the scavenging efficiencies by rain and snow
336 probably dominates the seasonal variations in sulfate. Reducing wet deposition by snow has been
337 found to improve model’s ability to reproduce the observed sulfate concentrations over the United
16 338 States (Paulot et al. 2016) and in the Arctic (Browse et al. 2012). The shift in the peak may be due
339 to the absence of an aging process since all sulfate is hydrophilic and the seasonal cycle of sulfate
340 production by SO2 oxidation.
341 The results discussed here may have implications for understanding the variability and trend in
342 Arctic aerosols and their climate impacts. Since large-scale circulation influences the key pro-
343 cesses of long-range transport and the spatial pattern of precipitation, natural climate variability at
344 annual to decadal timescales may play an important role in determining changes in aerosol con-
345 centrations in the Arctic (Christoudias et al. 2012; Eckhardt et al. 2003). It would be interesting
346 to apply our analysis to study the effect of climate variability on the characteristics of atmospheric
347 transport and wet deposition. As climate warms, precipitation at high latitudes is expected to in-
348 crease, but the fraction of snow may decrease (Barnett et al. 2005; Singarayer et al. 2006). These
349 changes tend to enhance the wet scavenging of aerosols and result in a cleaner Arctic atmosphere.
350 The aging process, which is affected by atmospheric composition, may also vary over time. SO2
351 emissions at mid-latitudes have generally declined in recent decades (Streets et al. 2006) and are
352 projected to decrease even more in the future (Levy et al. 2008). This long-term trend in SO2
353 emissions will result in a decrease in sulfate concentrations but an increase in BC concentrations
354 in the Arctic due to a slower aging process by condensation of H2SO4 and thus weaker wet re-
355 moval. Therefore it is important to consider the changes in aerosol sources and sinks when using
356 models to examine how aerosols may alter Arctic climate under future emission scenarios.
357 Acknowledgments. Songmiao Fan and Fabien Paulot provided helpful reviews of an ear-
358 lier draft. We acknowledge the Science & Technology Branch at Environment Canada, the
359 Global Monitoring Division at NOAA Earth System Research Laboratory, and Norwegian
360 Institute for Atmospheric Research for providing measurement data. The observations at
17 361 Alert, Barrow, and Zeppelin are available through the Canadian Aerosol Baseline Measure-
362 ment (CABM) Program Datasets (http://www.ec.gc.ca/donneesnatchem-natchemdata/
363 default.asp?lang=En&n=22F5B2D4-1), the ESRL/GMD ftp site (/ftp.cmdl.noaa.gov/
364 aerosol/brw/archive/), and World Data Center for Aerosol (EBAS) (http://ebas.nilu.
365 no). The HIPPO data (Wofsy et al. 2012) are available at the CDIAC HIPPO data archive
366 (http://hippo.ornl.gov/10secdownload/). Model simulations are archived at GFDL and
367 are available from the authors upon request.
368 References
369 Barnett, T. P., J. C. Adam, and D. P. Lettenmaier, 2005: Potential impacts of a warming climate
370 on water availability in snow-dominated regions. Nature, 438 (7066), 303–309, doi:10.1038/
371 nature04141.
372 Barrie, L. A., 1986: Arctic air pollution: An overview of current knowledge. Atmospheric Envi-
373 ronment (1967), 20 (4), 643–663, doi:10.1016/0004-6981(86)90180-0.
374 Bindoff, N., and Coauthors, 2013: Detection and Attribution of Climate Change: from Global
375 to Regional. Climate Change 2013 - The Physical Science Basis, Intergovernmental Panel
376 on Climate Change, Ed., Cambridge University Press, Cambridge, 867–952, doi:10.1017/
377 CBO9781107415324.022.
378 Bolin, B., and C. D. Keeling, 1963: Large-scale atmospheric mixing as deduced from the seasonal
379 and meridional variations of carbon dioxide. J. Geophys. Res., 68 (13), 3899–3920, doi:10.1029/
380 JZ068i013p03899.
381 Bond, T. C., and Coauthors, 2013: Bounding the role of black carbon in the climate system: A
382 scientific assessment. J. Geophys. Res., 118 (11), 5380–5552, doi:10.1002/jgrd.50171.
18 383 Browse, J., K. S. Carslaw, S. R. Arnold, K. Pringle, and O. Boucher, 2012: The scavenging
384 processes controlling the seasonal cycle in Arctic sulphate and black carbon aerosol. Atmos.
385 Chem. Phys., 12 (15), 6775–6798, doi:10.5194/acp-12-6775-2012.
386 Christoudias, T., A. Pozzer, and J. Lelieveld, 2012: Influence of the North Atlantic Oscillation on
387 air pollution transport. Atmos. Chem. Phys., 12 (2), 869–877, doi:10.5194/acp-12-869-2012.
388 Cozic, J., S. Mertes, B. Verheggen, D. J. Cziczo, S. J. Gallavardin, S. Walter, U. Baltensperger, and
389 E. Weingartner, 2008: Black carbon enrichment in atmospheric ice particle residuals observed
390 in lower tropospheric mixed phase clouds. J. Geophys. Res., 113 (D15), D15 209, doi:10.1029/
391 2007JD009266.
392 Donner, L. J., and Coauthors, 2011: The Dynamical Core, Physical Parameterizations, and Basic
393 Simulation Characteristics of the Atmospheric Component AM3 of the GFDL Global Coupled
394 Model CM3. J. Climate, 24 (13), 3484–3519, doi:10.1175/2011JCLI3955.1.
395 Eckhardt, S., and Coauthors, 2003: The North Atlantic Oscillation controls air pollution transport
396 to the Arctic. Atmos. Chem. Phys., 3 (5), 1769–1778, doi:10.5194/acp-3-1769-2003.
397 Eleftheriadis, K., S. Vratolis, and S. Nyeki, 2009: Aerosol black carbon in the European Arctic:
398 Measurements at Zeppelin station, Ny-Alesund,˚ Svalbard from 1998-2007. Geophys. Res. Lett.,
399 36 (2), doi:10.1029/2008GL035741.
400 Fan, S.-M., and Coauthors, 2012: Inferring ice formation processes from global-scale black carbon
401 profiles observed in the remote atmosphere and model simulations. J. Geophys. Res., 117 (D23),
402 D23 205, doi:10.1029/2012JD018126.
19 403 Friedman, B., G. Kulkarni, J. Beranek,´ A. Zelenyuk, J. A. Thornton, and D. J. Cziczo, 2011: Ice
404 nucleation and droplet formation by bare and coated soot particles. J. Geophys. Res., 116 (D17),
405 D17 203, doi:10.1029/2011JD015999.
406 Gallagher, M. W., 2002: Measurements and parameterizations of small aerosol deposition veloci-
407 ties to grassland, arable crops, and forest: Influence of surface roughness length on deposition.
408 J. Geophys. Res., 107 (D12), 4154, doi:10.1029/2001JD000817.
409 Garrett, T. J., S. Brattstrom,¨ S. Sharma, D. E. J. Worthy, and P. Novelli, 2011: The role of scav-
410 enging in the seasonal transport of black carbon and sulfate to the Arctic. Geophys. Res. Lett.,
411 38 (16), doi:10.1029/2011GL048221.
412 Hall, N. M. J., B. J. Hoskins, P. J. Valdes, and C. A. Senior, 1994: Storm tracks in a high-resolution
413 GCM with doubled carbon dioxide. Quart. J. Roy. Meteor. Soc., 120 (519), 1209–1230, doi:
414 10.1002/qj.49712051905.
415 Held, I. M., 1999: The macroturbulence of the troposphere. Tellus A, 51 (1 SPEC. ISS.), 59–70,
416 doi:10.3402/tellusa.v51i1.12306.
417 Iversen, T., and E. Joranger, 1985: Arctic air pollution and large scale atmospheric flows. Atmo-
418 spheric Environment (1967), 19 (12), 2099–2108, doi:10.1016/0004-6981(85)90117-9.
419 Janssens-Maenhout, G., and Coauthors, 2015: HTAP v2.2: a mosaic of regional and global emis-
420 sion grid maps for 2008 and 2010 to study hemispheric transport of air pollution. Atmos. Chem.
421 Phys., 15 (19), 11 411–11 432, doi:10.5194/acp-15-11411-2015.
422 Klonecki, A., P. Hess, L. Emmons, L. Smith, J. Orlando, and D. Blake, 2003: Seasonal changes in
423 the transport of pollutants into the Arctic troposphere-model study. J. Geophys. Res., 108 (D4),
424 8367, doi:10.1029/2002JD002199.
20 425 Law, K. S., and A. Stohl, 2007: Arctic air pollution: origins and impacts. Science, 315 (5818),
426 1537–40, doi:10.1126/science.1137695.
427 Levy, H., M. D. Schwarzkopf, L. Horowitz, V. Ramaswamy, and K. L. Findell, 2008: Strong sen-
428 sitivity of late 21st century climate to projected changes in short-lived air pollutants. J. Geophys.
429 Res., 113 (D6), D06 102, doi:10.1029/2007JD009176.
430 Lin, M., and Coauthors, 2012: Transport of Asian ozone pollution into surface air over the western
431 United States in spring. J. Geophys. Res., 117 (D21), doi:10.1029/2011JD016961.
432 Liu, J., S. Fan, L. W. Horowitz, and H. Levy, 2011: Evaluation of factors controlling long-range
433 transport of black carbon to the Arctic. J. Geophys. Res., 116 (D4), D04 307, doi:10.1029/
434 2010JD015145.
435 Mitchell, J. M., 1957: Visual range in the polar regions with particular reference to the Alaskan
436 Arctic. J. Atmos. Terr. Phys. Spec. Suppl, 195–211.
437 Newell, R. E., D. G. Vincent, and J. W. Kidson, 1969: Interhemispheric mass exchange from
438 meteorological and trace substance observations. Tellus A, doi:10.3402/tellusa.v21i5.10114.
439 Oshima, N., M. Koike, Y. Zhang, and Y. Kondo, 2009: Aging of black carbon in outflow
440 from anthropogenic sources using a mixing state resolved model: 2. Aerosol optical prop-
441 erties and cloud condensation nuclei activities. J. Geophys. Res., 114 (D18), D18 202, doi:
442 10.1029/2008JD011681.
443 Paulot, F., and Coauthors, 2016: Sensitivity of nitrate aerosols to ammonia emissions and to nitrate
444 chemistry: implications for present and future nitrate optical depth. Atmos. Chem. Phys., 16 (3),
445 1459–1477, doi:10.5194/acp-16-1459-2016.
21 446 Petters, M. D., A. J. Prenni, S. M. Kreidenweis, P. J. DeMott, A. Matsunaga, Y. B. Lim, and P. J.
447 Ziemann, 2006: Chemical aging and the hydrophobic-to-hydrophilic conversion of carbona-
448 ceous aerosol. Geophys. Res. Lett., 33 (24), L24 806, doi:10.1029/2006GL027249.
449 Quinn, P. K., G. Shaw, E. Andrews, E. G. Dutton, T. Ruoho-Airola, and S. L. Gong, 2007: Arctic
450 haze: current trends and knowledge gaps. Tellus B, 59 (1), 99–114, doi:10.1111/j.1600-0889.
451 2006.00238.x.
452 Scheuer, E., 2003: Seasonal distributions of fine aerosol sulfate in the North American Arctic
453 basin during TOPSE. J. Geophys. Res., 108 (D4), 8370, doi:10.1029/2001JD001364.
454 Schwarz, J. P., and Coauthors, 2013: Global-scale seasonally resolved black carbon vertical pro-
455 files over the Pacific. Geophys. Res. Lett., 40 (20), 5542–5547, doi:10.1002/2013GL057775.
456 Sharma, S., 2004: Long-term trends of the black carbon concentrations in the Canadian Arctic. J.
457 Geophys. Res., 109 (D15), D15 203, doi:10.1029/2003JD004331.
458 Sharma, S., E. Andrews, L. a. Barrie, J. a. Ogren, and D. Lavoue,´ 2006: Variations and sources of
459 the equivalent black carbon in the high Arctic revealed by long-term observations at Alert and
460 Barrow: 19892003. J. Geophys. Res., 111 (D14), D14 208, doi:10.1029/2005JD006581.
461 Shaw, G., 1981: Eddy diffusion transport of Arctic pollution from the mid-latitudes: A prelimi-
462 nary model. Atmospheric Environment (1967), 15 (8), 1483–1490, doi:10.1016/0004-6981(81)
463 90356-5.
464 Shindell, D., and G. Faluvegi, 2009: Climate response to regional radiative forcing during the
465 twentieth century. Nat. Geosci., 2 (4), 294–300, doi:10.1038/ngeo473.
466 Shindell, D. T., and Coauthors, 2008: A multi-model assessment of pollution transport to the
467 Arctic. Atmos. Chem. Phys., 8 (17), 5353–5372, doi:10.5194/acp-8-5353-2008.
22 468 Singarayer, J. S., J. L. Bamber, and P. J. Valdes, 2006: Twenty-First-Century Climate Impacts from
469 a Declining Arctic Sea Ice Cover. J. Climate, 19 (7), 1109–1125, doi:10.1175/JCLI3649.1.
470 Stohl, A., 2006: Characteristics of atmospheric transport into the Arctic troposphere. J. Geophys.
471 Res., 111 (D11), D11 306, doi:10.1029/2005JD006888.
472 Stohl, A., Z. Klimont, S. Eckhardt, K. Kupiainen, V. P. Shevchenko, V. M. Kopeikin, and
473 a. N. Novigatsky, 2013: Black carbon in the Arctic: the underestimated role of gas flar-
474 ing and residential combustion emissions. Atmos. Chem. Phys., 13 (17), 8833–8855, doi:
475 10.5194/acp-13-8833-2013.
476 Streets, D. G., Y. Wu, and M. Chin, 2006: Two-decadal aerosol trends as a likely explanation of
477 the global dimming/brightening transition. Geophys. Res. Lett., 33 (15), L15 806, doi:10.1029/
478 2006GL026471.
479 Wiedinmyer, C., S. K. Akagi, R. J. Yokelson, L. K. Emmons, J. a. Al-Saadi, J. J. Orlando, and
480 a. J. Soja, 2011: The Fire INventory from NCAR (FINN): a high resolution global model to
481 estimate the emissions from open burning. Geosci. Model Dev., 4 (3), 625–641, doi:10.5194/
482 gmd-4-625-2011.
483 Wofsy, S. C., 2011: HIAPER Pole-to-Pole Observations (HIPPO): fine-grained, global-scale mea-
484 surements of climatically important atmospheric gases and aerosols. Philos. Trans. R. Soc., A,
485 369 (1943), 2073–2086, doi:10.1098/rsta.2010.0313.
486 Wofsy, S. C., and Coauthors, 2012: HIPPO Merged 10-second Meteorology, Atmo-
487 spheric Chemistry, Aerosol Data. (R 20121129). Carbon Dioxide Information Analy-
488 sis Center, Oak Ridge National Laboratory, Oak Ridge, Tennessee, U.S.A. htt p :
489 //dx.doi.org/10.3334/CDIAC/hippo 010 (Release20121129).
23 490 Zhang, R., A. F. Khalizov, J. Pagels, D. Zhang, H. Xue, and P. H. McMurry, 2008: Variability
491 in morphology, hygroscopicity, and optical properties of soot aerosols during atmospheric pro-
492 cessing. Proc. Natl. Acad. Sci. (USA), 105 (30), 10 291–10 296, doi:10.1073/pnas.0804860105.
24 493 LIST OF FIGURES
494 Fig. 1. (top) Model simulated and observed monthly mean surface BC at Alert (2008-2012), Bar- 495 row (2008-2013), and Zeppelin (2008-2013). Error bars denote one standard deviation from 496 monthly means. (bottom) Model simulated and observed BC vertical profiles at high lat- ◦ ◦ 497 itudes (66 -85 N) during HIPPO (HIPPO1: January 2009, HIPPO2: October-November 498 2009, HIPPO3: March-April 2010, HIPPO4: June-July 2011, HIPPO5: August-September 499 2011). Error bars associated with observational profiles represent one standard deviation 500 of the HIPPO data. For the comparison, daily BC fields archived from the model are first ◦ ◦ 501 sampled along the flight track and then averaged over 66 -85 N for each campaign. . . . . 26
502 Fig. 2. Monthly mean (a) Arctic BC column burdens simulated using AM3 (red) and derived from 503 the box model (black), (b) AM3 simulated total meridional BC flux into the Arctic (solid) 504 and contributions from the mean meridional circulation (dashed), stationary eddies (dash- 505 dot), and transient eddies (dotted). Error bars denote one standard deviation from monthly 506 means...... 27
◦ 507 Fig. 3. Vertical profiles of meridional BC flux at 66 N. The solid and dashed lines represent DJF 508 and JJA, respectively...... 28
509 Fig. 4. Time-filtered meridional BC flux ({vncn} defined in Section 5) as a fraction of total flux at ◦ 510 66 N. The solid and dashed lines represent DJF and JJA, respectively...... 29
511 Fig. 5. Scatter plot of vertically integrated monthly mean meridional BC flux by transient eddies ◦ ◦ 512 versus meridional gradients of BC column burden averaged over 40 -66 N. The linear re- 513 gression line is also shown. Green and red dots represent DJF and JJA, respectively. . . . 30
514 Fig. 6. Monthly mean AM3 simulated (a) hydrophilic fractions of Arctic BC (solid) and BC flux ◦ ◦ ◦ ◦ 515 into the Arctic (dashed), (b) BC aging time averaged at 66 -90 N (solid) and 40 -66 N 516 (dashed), (c) Arctic BCpi wet deposition efficiency, and (d) rate coefficients of conversion 517 of cloud condensate to rain (solid) and snow (dashed) in the Arctic. Error bars denote one 518 standard deviation from monthly means...... 31
25 HIPPO1 HIPPO2 HIPPO3 HIPPO4 HIPPO5 100 JAN 2009 Oct-Nov 2009 Mar-Apr 2010 Jun-Jul 2011 Aug-Sep 2011
300
500
Pressure(hpa) Obs 700 AM3 900 0.1 1 10 100 0.1 1 10 100 0.1 1 10 100 0.1 1 10 100 0.1 1 10 100 BC (ng kg−1 )
519 FIG. 1. (top) Model simulated and observed monthly mean surface BC at Alert (2008-2012), Barrow (2008-
520 2013), and Zeppelin (2008-2013). Error bars denote one standard deviation from monthly means. (bottom) ◦ ◦ 521 Model simulated and observed BC vertical profiles at high latitudes (66 -85 N) during HIPPO (HIPPO1: Jan-
522 uary 2009, HIPPO2: October-November 2009, HIPPO3: March-April 2010, HIPPO4: June-July 2011, HIPPO5:
523 August-September 2011). Error bars associated with observational profiles represent one standard deviation of
524 the HIPPO data. For the comparison, daily BC fields archived from the model are first sampled along the flight ◦ ◦ 525 track and then averaged over 66 -85 N for each campaign.
26 (a) (b)
526 FIG. 2. Monthly mean (a) Arctic BC column burdens simulated using AM3 (red) and derived from the box
527 model (black), (b) AM3 simulated total meridional BC flux into the Arctic (solid) and contributions from the
528 mean meridional circulation (dashed), stationary eddies (dash-dot), and transient eddies (dotted). Error bars
529 denote one standard deviation from monthly means.
27 (a)
◦ 530 FIG. 3. Vertical profiles of meridional BC flux at 66 N. The solid and dashed lines represent DJF and JJA,
531 respectively.
28 ◦ 532 FIG. 4. Time-filtered meridional BC flux ({vncn} defined in Section 5) as a fraction of total flux at 66 N. The
533 solid and dashed lines represent DJF and JJA, respectively.
29 8 DJF 7 JJA ) -1 6 (kg s (kg 5
BC flux by transient eddies by BCflux 4 −3.5 −3.0 −2.5 −2.0 1e−6 BC column gradient (kg m-2 )
534 FIG. 5. Scatter plot of vertically integrated monthly mean meridional BC flux by transient eddies versus ◦ ◦ 535 meridional gradients of BC column burden averaged over 40 -66 N. The linear regression line is also shown.
536 Green and red dots represent DJF and JJA, respectively.
30 (a) (b) 1.0 25 66N-90N 20 0.8 40N-66N 15 0.6 10 0.4 burden aging (day) time
hydrophilicfraction 5 flux 0.2 0 J F M A M J J A S O N D J F M A M J J A S O N D
) (c) (d) -1 0.20 30 )
-1 rain 25 0.16 snow 20 0.12 15 10 0.08 5 rate coefficientrate (day 0.04 0
wet depositionwet efficiency (day J F M A M J J A S O N D J F M A M J J A S O N D
537 FIG. 6. Monthly mean AM3 simulated (a) hydrophilic fractions of Arctic BC (solid) and BC flux into the ◦ ◦ ◦ ◦ 538 Arctic (dashed), (b) BC aging time averaged at 66 -90 N (solid) and 40 -66 N (dashed), (c) Arctic BCpi wet de-
539 position efficiency, and (d) rate coefficients of conversion of cloud condensate to rain (solid) and snow (dashed)
540 in the Arctic. Error bars denote one standard deviation from monthly means.
31